Background Several studies suggest the gut microbiome may be a novel, modifiable biomarker for clinical efficacy of immune checkpoint inhibitors (ICIs). Microbiome profiling of pre-treatment samples demonstrated that high alpha-diversity and enrichment of specific bacterial species are associated with improved tumor responses in melanoma, renal cell cancer (RCC), and non-small cell lung cancer (NSCLC). Despite these reports, the specific bacteria or communities helpful or harmful have been inconsistent across study populations, and further correlation with immune and mutational biomarkers are limited or lacking. We hypothesize that, by use of a larger sample size and a consistent computational approach, we would derive a clearer microbial profile that correlated with immunotherapeutic outcomes.

Methods We re-analyzed the available raw 16S rRNA amplicon and metagenomic sequencing data across five recently published ICI studies (n=303 unique patients) of responder (R) and nonresponse (NR) using a consistent computational approaches (Resphera Insight and MetaPhlAn2). Using novel microbiota signatures, we identified Re-analysis Indices for R- and NR-associated bacteria and validated the result in three addition cohorts with available raw sequencing data in patients with melanoma, hepatocellular cancer (HCC), and NSCLC (n=105).

Results Our results confirm signals reported in each study, though some bacteria reported initially were not statistically significant after correction for false discovery rate. Likely, in part, because our analysis allows for comparison of individual species across cohorts, we were able to identify new bacterial signatures, such as Oxalobacter formigenes, Roseburia hominis and Veillonella parvula, Clostridium hathewayi, enriched in R and NR respectively. When our Re-analysis Index was compared to an index assembled from the literature, we noted improvement occurred in a sensitivity and specificity analysis, especially in NR-associated signals. Moreover, we found that alpha-diversity was not consistently predictive of response or nonresponse to ICIs. Our Re-analysis Index also validated in melanoma patients and HCC but did not perform as well in the NSCLC cohort, suggesting the need for further refinement based on tumor type.

Conclusions In summary, this bioinformatics platform improves on existing pipelines by standardizing critical preprocessing and downstream analysis tools, enabling comprehensive evaluations of taxonomic and functional signals across sequencing datasets. Notably, the NR-associated Re-analysis Index shows the strongest and most consistent signal using a random effects model and in a sensitivity and specificity analysis (p < 0.01). Our integrated analyses suggest an approach to identify patients who would benefit from microbiome-based interventions targeted to improve response rates by using a biomarker for nonresponse.